Applying Large Language Models to Medical Papers

Fine-tuning large language models (LLMs) for
domain-specific tasks can significantly enhance their
performance in knowledge extraction and summarization. Biomedical literature is vast and often fragmented, making it challenging to efficiently synthesize relevant information.

Our project was built upon Dr. Rojas' research on natural language processing (NLP) in biomedical and bioinformatic applications. We proposed to fine-tune Meta’s recently released Llama-3 model for enhanced relation extraction and summarization within biomedical contexts. Utilizing San Jose
State University’s high performance computing cluster, our approach leveraged domain-specific datasets to significantly
improve the efficiency of relation extraction tasks. The model’s performance was benchmarked against
state-of-the-art (SOTA) models, such as BioBERT and GPT-4, to evaluate its effectiveness in extracting meaningful insights. These results aim to contribute to pharmaceutical innovation and public health by reducing the time dedicated to manual literature reviews and identifying novel
drug-disease relationships.